More from the "how we did this?" section. ("This", being "fabricated the result we wanted with bigger numbers for the republicans")

Accordingly, like most political science work that employs DW-NOMINATE scores, this analysis focuses on the primary liberal/conservative scale. That scale runs from -1 (most liberal) to 1 (most conservative). Each lawmaker is assigned a value between those endpoints based on their voting record; the scores are designed to be comparable between Congresses and across time.

Literally no fixed or objective measure of values. Just everything measured relative to whoever is arbitrarily selected as the "most" lib/con for that given year and also blindly assuming they are equidistant from the center. You could just as well look at that data and say the democrats extremes are not lessening or even growing more extreme, whilst republican extremes are becoming more and more moderate with time.